Stay Informed
Get the latest news and ideas from Wallace.

1.2 Shared Leadership: Effects on Teachers and Students of Principals and Teachers Leading Together - Learning From Leadership


Learning From Leadership: Investigating the Links to Improved Student Learning

Click here to download the full report:
 Learning From Leadership: Investigating the Links to Improved Student Learning

Key Findings

  • Leadership practices targeted directly at improving instruction have significant effects on teachers‘ working relationships and, indirectly, on student achievement.
  • When principals and teachers share leadership, teachers‘ working relationships are stronger and student achievement is higher.
  • Leadership effects on student achievement occur largely because effective leadership strengthens professional community—a special environment within which teachers work together to improve their practice and improve student learning. Professional community, in turn, is a strong predictor of instructional practices that are strongly associated with student achievement.
  • The link between professional community and student achievement may be explained by reference to a school climate that encourages levels of student effort above and beyond the levels encouraged in individual classrooms.
  • Students in elementary schools perform better on state tests than students in upper grades. Principal leadership practices are unable, by themselves, to overcome this difference.
  • The factor of trust is less significant than the factors of instructional leadership and shared leadership (although it is associated with both).


Section 1.1 describes the extent to which a wide array of stakeholders may influence school decisions; it also describes the effects of broadly based influence on student learning. Section 1.2 focuses more narrowly on relationships among actors within schools, examining leadership shared by principals and teachers as it may affect classroom practice and student learning.

We focus here on principals and teachers for two main reasons. First, professionals acting within schools are uniquely well positioned to affect students‘ classroom experience. Second, the narrower focus pushes us beyond a simple definition of leadership as influence, to a more explicit specification of the functions responsible for such influence.

Section 1.2 addresses two questions:

  • Do three specific attributes of principals‘ leadership behavior—the sharing of leadership with teachers, the development of trust relationships among professionals, and the provision of support for instructional improvement—affect teachers‘ work with one another, and their classroom practices?
  • Do these leadership behaviors and attributes contribute to student achievement?

Prior Evidence

Prior evidence relevant to this component of our study identifies factors related to shared leadership, school conditions mediating the effects of shared leadership, and effective classroom instruction. We focus on variables that may contribute to a school‘s culture and climate, including (1) variables on which principals can have some direct effect, such as principal-teacher relations, trust, and shared leadership; (2) variables on which principals may have less influence, such as teacher-to-teacher relations in professional communities, and collective responsibility; and (3) variables on which the principal has indirect control, such as teachers‘ sense of personal efficacy, and the quality of instruction.

We assume that the effects of principal leadership on students are almost entirely indirect.54 The long line of research on school effectiveness shows that classroom environment and the quality of instruction are the variables linked most strongly to student learning (although some questions remain about the relative effectiveness of specific modes of instruction).55 Teacher characteristics (such as type of degree or certification) have limited effects,56 operating for the most part indirectly, through their impact on instruction.57 In other words: to learn how leadership contributes to student learning, we must ask how leadership affects instruction.58

Starting with Instruction

Various models of good instruction have evolved over the last several decades, but differences among them remain only partially resolved. An early review of research showed that certain instructional practices—e.g., using academic objectives to establish learning expectations, using particular strategies for classroom management, and pacing instruction appropriately, given the content to be taught and the characteristics of the learners—were consistently associated with student achievement.59 After the late 1980s, theory and research increasingly emphasized inquiry-based instructional models, in which the teacher‘s most important role was in designing lessons or learning experiences that involved guiding students toward new understanding through exploration and induction.60 While some approaches to constructivism emphasized modest roles for teachers (as "guides on the side"), others gave teachers clear responsibilities consistent with traditional roles, but also for organizing learning environments that develop students‘ sense of responsibility for their own learning.61

Researchers today rarely address "time on task" as a simplistic factor. Still, a growing body of evidence shows that student learning is enhanced when teachers exercise appropriate control over the pacing of classroom work,62 at least when the activity in question is based on rich materials and stimuli. Recent reviews have begun to reemphasize the role of the teacher in directing student learning.63

A particular problem is that research based on observations of instruction in widely varying settings (e.g., different disciplines, different grade levels) often yields little in the way of details sufficiently specific to understand the choices particular teachers must make.64 Taking adequate account of the complexity of classroom instruction is very difficult. As Cohen, Raudenbush, and Ball (2003) note, this is because teachers and students are independent and idiosyncratic actors. What happens instructionally in a given situation is context-specific, making it difficult to generalize validly about particular reform efforts aimed, for example, at developing shared leadership and professional community. Moreover, research to date has done little to identify direct links between the policies and practices of school-level leaders and the provision of high-quality instruction, whether teacher-directed or teacher-guided.

In a previous paper we used factor analysis to demonstrate that teachers report a distinctive style of teaching—one that incorporates direct influence over the pacing and content of classroom work while also providing opportunities for students to take charge of their own learning and construct their own knowledge. We called this style of teaching "focused instruction."65 In our view, if we overlook certain teacher-educator debates,66 our finding that "real teachers" combine elements of a traditional teacher-centered model with elements of constructivist models is consistent with other research on instructional approaches that are linked to student achievement.67

Instructional Leadership

As Hallinger (2005) notes in a recent review of scholarship, instructional leadership is an idea that refuses to go away, although it has been poorly defined since it was first introduced in the 1970s. In the school building, the principal is expected to understand the tenets of quality instruction, and to have sufficient knowledge of the curriculum to ensure that appropriate content is being delivered to all students.68 This presumes that the principal is capable of providing constructive feedback to improve teaching, or that she or he can design a system in which others provide this support. Research shows that consistent, well-informed support from principals makes a difference,69 and principals accordingly face increasing pressure to deliver (or at least promote) better support for instruction.

In their efforts to act as instructional leaders, some principals benefit from support provided, for example, through professional development programs; those who do are more likely than others to enact this leadership role consistently.70 While some scholars emphasize the importance of principals‘ deep understanding of curricular content and instructional materials,71 others pay more attention to principals‘ support for improved instruction.72 Typically, those who emphasize the importance of deep content knowledge study elementary schools.73 Even in elementary schools, however, the principal‘s ability to provide support through effective interaction may be more important than his or her specific content knowledge.74

Middle and high school principals cannot be expected to provide substantive support for instruction, given the multiple disciplines that are taught in their schools. Thus, many studies of instructional leadership in secondary schools emphasize the development of improved learning environments for teachers, focusing on the ability of principals to stimulate teachers‘ innovative behavior.75 Because our study includes secondary schools, we chose to emphasize supportive behaviors as well as direct coaching or modeling of instruction.

Shared Leadership

For more than three decades, reform proposals have recommended the inclusion of teachers in shared leadership roles. In the late 1980s and early 1990s, efforts to promote school-based management often included formal representation of teachers in decision making—although many investigations of these efforts report weak implementation.76 Recent policy discussions (within, e.g., the Education Commission of the States, the Council of Chief State School Officers, and teacher professional associations) suggest broad support now for expanding teachers‘ participation in leadership and decision-making tasks. These discussions are compatible with findings from some research which suggests that increasing teacher influence may improve schools significantly.77 Other research, however, suggests that teachers‘ involvement in formal decision-making or leadership roles will have limited impact on student achievement.78

Still, what constitutes and promotes the distribution or sharing of leadership in a school is somewhat unclear. Sharing leadership may have its greatest impact by reducing teacher isolation and increasing commitment to the common good.79 Experiencing informal influence and feedback through professional discussions encourages a focus on shared practices and goals,80 and it may foster organizational innovation.81 In this paper we define shared leadership broadly to denote teachers‘ influence over, and their participation in, school-wide decisions with principals. This view of shared leadership reflects an emerging consensus among scholars about the people who are concerned with formal and informal enactments of leadership roles; it also distinguishes our approach from the approach of scholars who blend the concept of shared leadership with instructional leadership.82


The concept of organizational trust has been a staple of organizational research for some time. It matters a great deal whether participants in an organization trust the decision-making capacity of the organization‘s leaders. Driscoll (1978) found that such trust predicts overall satisfaction with the organization better than employee participation in decision making. A more recent study examined changes in levels of trust within work teams; it found that the perceived ability of colleagues was a strong predictor of trust, and that trust was a significant predictor for risk-taking behaviors.83

In the past two decades, studies of trust as a factor in school improvement have begun to illuminate certain actions leaders take to alter the culture in a school positively.84 In a sample of secondary schools, Tarter et al. (1989) found that supportive principal behavior and faculty trust were significantly correlated. In schools with higher levels of engaged teachers, moreover, teachers expressed higher levels of trust in their colleagues. Tarter‘s study implies that principals can build trust indirectly through supportive behavior, but they cannot make teachers trust one another through direct action. Similarly, Bryk and Schneider‘s (2003) study of Chicago elementary schools found that principal respect and personal regard for teachers, competence in core role responsibilities, and personal integrity were associated with relational trust among all adult members of the school. Louis (2007) identified similar principal behaviors that affect trust, and also linked trust to shared leadership. High-trust schools exhibited more collective decision making, with a greater likelihood that reform initiatives were widespread, and with demonstrated improvements in student learning. Tschannen-Moran also outlined key leadership behaviors and specific actions that engender trust. For example, "Competence" is enacted by "engaging in problem solving, setting standards, buffering teachers, pressing for results" (2004, p.34). More recently, trust has been shown to predict how educators interpret their superiors‘ ability to carry out more technical and transformational leadership functions.85

Embedded in the notion of trust is the key distinction between the "trustee" and the "trustor,"—that is, those having more or less power (or dependence) in a particular situation.86 Teachers‘ views of trustworthy principals tend to be based on the leadership characteristics outlined above. However, we have much less information about why principals do or do not trust their teachers.

Teacher Leadership and Professional Community

While we have focused thus far on shared leadership and principal-teacher trust, teacher-teacher relationships are even more important as a foundation for the way in which teachers work to improve instruction,87 and how they are affected by the leadership behavior of principals.88 Here we emphasize the importance of professional community, largely because accumulating evidence shows that it is related to improved instruction, student achievement,89 and one of our leadership variables (shared leadership).90

York-Barr and Duke (2004) view professional community as a vehicle for the exercise of teacher leadership, a perspective that we adopt in this paper. Supportive interaction among teachers in school-wide professional communities enable them to assume various roles with one another as mentor, mentee, coach, specialist, advisor, facilitator, and so on. However, professional community amounts to more than just support; it also includes shared values, a common focus on student learning, collaboration in the development of curriculum and instruction, and the purposeful sharing of practices—all of which may be thought of as distributed leadership.91

Findings from several studies cited above suggest that when the professional community focuses on the quality of student learning, teachers adopt instructional practices that enhance students‘ learning. While many factors affect whether or not professional community exists in a school, one highly significant factor is strong leadership by principals.92 Professional community is closely associated with organizational learning, and the term "professional learning communities" has become a common shorthand expression among practitioners. Thus, the presence of a professional community appears to foster collective learning of new practices—when there is principal leadership.93

School Level

Many characteristics of schools may moderate leadership effects. In this paper, we focus on potential differences between elementary and secondary schools. Investigations of principal leadership effects on teachers and students are often carried out using only one type of school.94 Those that use samples from all levels (e.g., Marks & Printy, 2003) are based on a small number of cases, while those based on a larger number of schools often use a convenience sample drawn from a single district.95 Nevertheless, there is reason to suppose that leadership practices and their effects may be different in elementary and secondary schools, given differences of school size and organization. The principal in a very large school simply does not have time to work directly with all teachers. As Harris (2002) points out, secondary school principals seem to influence teachers and teaching practice because of the organizational climate they create, not through specific interactions or interventions.

New Evidence

An analytic framework derived from prior scholarship and our previous investigation of the relationship between principal leadership and instruction guided our examination of shared leadership.96 We assumed that both principal-teacher relationships (indicated by trust, instructional leadership, and perceptions of shared leadership) and teacher-teacher relationships (indicated by professional community) will affect classroom practice. Classroom practice—particularly the type of instruction that combines elements of teacher-directed and constructivist approaches—should, in turn, affect student learning. We emphasize the importance of classroom practice as the direct cause of increased student learning because there is little evidence, from either survey or qualitative research, that principal leadership can have a direct effect apart from changes in teacher practice.

Our specific intention, once again, is to explore two questions:

  • Do three specific attributes of principals‘ leadership behavior—the sharing of leadership with teachers, the development of trust relationships among professionals, and the provision of support for instructional improvement—affect teachers‘ work with each other and their classroom practices?
  • Do these leadership behaviors and attributes contribute to student achievement?


This component of our study utilized data from the first and second round of teacher surveys. Each of the two surveys contained some items from established instruments, as well as many new items. This section of our study is based on surveys of 4,491 teachers in 43 districts in 157 schools, with a response rate of 67% (for Round One, in 2005), and 3,900 teachers in 40 districts in 134 schools, with a response rate of 55% (for Round Two, in 2008).97 It combines some measures from the first teacher survey (principal leadership variables) with some from the second teacher survey (measures of trust, and an improved measure of focused instruction).

We measured each of the variables in the surveys using multiple items. The items and their alphas are shown in Table 1.2.1. We wish to draw particular attention to the Focused Instruction variable, which combines elements of constructivist (student controlled) and direct (teacher controlled) behaviors. The other measures are based largely on items that we adapted for this study from previous surveys.

Using the conceptual framework outlined above, we initially performed correlation analyses and stepwise linear regressions. We then used causal modeling (using the SPSS AMOS program) to examine the direct and indirect effects of leadership on achievement. We chose mathematics achievement as our dependent measure largely because within-school variability in instructional quality may be lower for mathematics than for other subjects.98 However, we also conducted comparable analyses using state literacy test scores, with results similar to those reported below.

The Indirect Nature of Leadership Effects

We initially assumed that the effects of leadership on student achievement are largely indirect, operating through other variables. We examined this assumption by examining correlations, which are presented in Table 1.2.2. The results indicate that achievement scores in mathematics are significantly associated with focused instruction, professional community, and teachers‘ trust in the principal; they are not significantly associated with principal behaviors (instructional leadership and shared leadership), which provides support for our assumption. Trust in the principal and professional community, on the other hand, are both associated with achievement in mathematics, which suggests that relationships among adults may be important factors determining how well students perform. In our sample, students in elementary schools perform better than students in secondary schools on state benchmark tests.

If we look at the remaining cells in the correlation matrix, it is clear that the measures of predictors are highly correlated. Our data are consistent with results from other studies in suggesting, for example, that on many measures the quality of teachers‘ work life (trust, professional community, experience of strong leadership) is lower in secondary schools.99 In addition, teachers whose experience with other adults is positive on one of our dimensions tend to have similarly positive responses on the others. In sum, while the results are confirmatory, they suggest a need for further analysis to investigate how the relationships among the variables may combine to affect teachers‘ classroom practice and student learning. We therefore conducted several stepwise regression analyses to address the two questions serving as the focus for this sub-study.

Effects on Teachers’ Work of Selected Attributes of Leadership Behavior

To address this question, we performed further analyses on results from our earlier investigations,100 looking at the relationship between principal behaviors and characteristics and teachers‘ instructional practice. The results of this regression are presented in Table 1.2.3.

Using a 3 model approach, we first examined the relationship between professional community and focused instruction, adding principal behaviors and characteristics in model 2, and finally adding school level, which has been shown in previous studies to affect both professional community and instruction. The results suggest that professional community and trust in the principal are the only significant predictors. In addition, until building level is added in model 3, professional community seems to bear more weight than trust (the change in the relationship in model 3 is presumably accounted for by the negative relationship between being a secondary school and trusting the principal). It is apparently the case that collegial relationships among adults in the school, whether principal-teacher or teacher-teacher, lead to stronger focused instruction.

The Influence of Principal Leadership on Student Achievement

To address the second question, about the effects of principal leadership on student achievement, we again used a 3 model approach.

We looked first at the instruction-learning relationship in model 1, then added professional community (teacher-teacher relationships) as a second step, and finally added both building level and leadership characteristics in a third stage (Table 1.2.4). The results indicate that instructional practices have a significant effect on achievement (Model 1), but that this effect is diminished when we introduce teachers‘ professional community (Model 2), and it is further diminished when we look at school level and school demographic characteristics (Model 3).

The second regression model shows that adding professional community to the simple instruction-achievement model barely raises the percentage of variance explained. However, when the leadership variables are added in model 3, there is a large increase in the R and R2, which suggests that principal leadership, even if it operates indirectly, is important. Both trust in leadership and instructional leadership exhibit significant regression coefficients, while building level and shared leadership are insignificant. Overall, adding leadership variables and the building level control variable more than double the percentage of explained variance in mathematics achievement. In other words, the regression evidence is strong for a relatively important leadership effect.

While the regressions support our assumption that leadership affects student learning, we assumed that it was unwise to over-interpret the regression coefficients, given the relatively high correlations among the predictor variables. In addition, the results of the two regressions raise as many questions as they answer. Why, for example, does instructional leadership exercised by principals have an insignificant effect in the regressions that focused on instruction as the dependent variable, while it shows a strong effect when the dependent variable is student achievement? We therefore moved to test our assumptions through causal modeling, guided by a set of possible interpretations of the regressions, as well as the literature reviewed above.

Figure 4 presents the model that illustrates the least complicated approach to answering the two questions motivating our inquiry.

Figure 4: Effects of Principals’ Leadership Behavior on Teachers and Student Achievement

The model makes the simplifying assumption that we do not know enough to examine a causal relationship among the three measures of leadership behavior/characteristics. They are, thus, positioned, along with the dichotomous variable reflecting the building level (elementary/secondary) at the left side of the model. In light of prior research, we then assume that leadership behaviors and characteristics are the factors most likely to create the conditions for professional community to develop among teachers. We discuss additional assumptions in our interpretation of results, which follows.

We used the maximum likelihood method for the path analysis. We assessed goodness of fit between the model and the data via three fit indices: the goodness-of-fit index (GFI), normed fit index (NFI), and the comparative fit index (CFI). GFI, NFI, and CFI values greater than .9 indicate that the model is a good fit with the data.101 The CFI is particularly critical, since it is a useful statistic with relatively small samples.102 The values of the GFI (.952), CFI (.924,), and the NFI (.900) all meet the suggested criterion. Taken together, these results indicate that the fit between the model and the data is adequate.103

We interpret the findings of the path analysis as follows:

  • Although principal instructional leadership has significant effects on teachers‘ working relationships (professional community), its direct effects on instruction are limited.
  • Shared leadership was not assumed to have a direct effect on instruction, but rather an indirect effect through professional community as a locus for teacher leadership focused on instructional improvement. The model confirms this indirect relationship.
  • Trust, which represents the emotional bond between the principal and teachers, was assumed to have a strong impact on teacher-teacher relationships. The model suggests, however, that its impact on professional community is limited, compared to the effect of leadership behaviors.
  • Building level, as expected, has a strong effect on professional community (with elementary schools being advantaged), and an equally strong direct effect on achievement (again, an elementary school advantage), but no significant effect on focused instruction. We did not expect the latter result; it suggests a need for further investigation to clarify the dynamics of professional community, instruction, and achievement in high schools. We explore this topic further in Section 1.5.
  • Professional community has significant indirect effects on achievement, owing to its strong relationship to focused instruction.


Efforts to determine how principal leadership affects student achievement have a rich, albeit recent, history. Our analysis provides the most extensive empirical test to date of whether instructional leadership, shared leadership, and trust in the principal, when considered together, have the potential to increase student learning. The answer is an unqualified yes, but the findings are complex and suggest a need for further analysis.

First, the emotional side of principal behavior—which we have assessed by reference to teachers‘ trust in the principals as ethical, caring, and competent—has on its own been shown to have a strong relationship to student outcomes. In our study, however, its relative significance diminishes when we take into consideration principal behaviors, as measured by our constructs of instructional leadership and shared leadership. Still, we are not prepared, based on a single study and a simple path model, to discount the importance of the emotional side of leadership, which has been shown in studies in industry as well as education to have powerful effects on the way in which people engage with their work.

Because trust is highly correlated with other key measures used in this study, we are inclined to say that our assumption—that trust is not reciprocal, for example, with professional community—is unwarranted. Further investigation is warranted to determine how the emotional side of leadership interacts with other leadership behaviors and with teachers‘ relationships with one another. Follow-up research might build on existing work, but it also should attend more directly to instrumental leadership actions.104

Shared leadership and instructional leadership are important variables, but they are indirectly related to student achievement. Both seem to gain their influence because of their strong relationships to other variables: to the way in which teachers organize themselves into professional communities, to reflective discussions about instruction, and to a sense of collective responsibility for student learning. This finding is hardly surprising when we consider the arguments for shared leadership, which generally emphasize expanding the sphere of responsibility and creativity to meet pressing school needs. The largely indirect effects of instructional leadership are, however, equally significant. While principals may engage in classroom visits and model good teaching by working with individual teachers, individual interventions (which would have emerged as a direct effect on good classroom practice) seem less important than detailed investigations of elementary schools suggest.105

The finding is important because shared leadership and instructional leadership are often regarded as alternative strategies for reaching the desired end of student learning. Those advocating instructional leadership emphasize the need to maintain a singular focus on classroom practice as the key to improving student achievement, and they point to the important role of the principal as a model. Others who look at shared leadership point to the importance of creating a learning organization in which all eyes are focused on leadership for learning. Our data suggest that these are complementary approaches, and that both may be necessary. Thus, using a larger and more diverse sample, we affirm Marks and Printy‘s (2003) work, which emphasizes the importance of combining leadership foci (in their case, transformational and instructional).

The findings regarding differences between elementary and secondary schools are particularly important as we begin to develop theories of effective school leadership. Our results, as we have noted, suggest the need for further inquiry; still, it is clear that the job of fostering student achievement is far easier in elementary schools than in secondary schools.

Implications for Policy and Practice

Four implications for policy and practice emerged from this section of our study.

  1. Teachers and educators holding formal administrative responsibilities need to acknowledge and act on the importance of collective, shared efforts to improve instruction.

    Professional community is regarded by some teachers as a code term for an administratively initiated program designed to encourage teachers to analyze student achievement data and turn it into improved test scores. Our analysis suggests that the reality is more complex. Teachers do need to work together to improve instruction and student learning, but administrators also need to be part of the process. The process may be as simple as having principals participate in professional development activities for teachers, or as complex as reorganizing the formal authority structure of the school. In any case, it requires a rethinking of the "bright line" that often separates administration and teaching.

  2. To realize their potential as instructional leaders, principals working in middle schools and high schools need particular modes of support. They face a distinct challenge, shaped by the large, complex settings in which they work, and the level of support extended to them should be commensurate with their distinct needs.

    Simply increasing the pressure on principals is unlikely to bring about real improvements in principal-teacher collaboration and achievement levels in secondary schools. Many school districts, however, lack the capacity to do more than that. We suggest accordingly that entities at the state or the regional/national level will need to be involved. Because we know from international studies ( PISA and TIMSS, e.g.) that secondary schools are the weakest link in our educational system, and that they show limited capacity for improvement under current accountability policies, we suggest that designing and providing new programs to support secondary school principals must become a policy priority.

  3. Principal preparation and professional development programs should continue to emphasize both the "softer" (emotional) and the "harder" (behavioral) aspects of leadership. While our results suggest that principals‘ behavior is more important than the levels of trust principals evoke, behavior and levels of trust are empirically part of a bundle that is difficult to disentangle. Trust without instructional and shared leadership to support it may be of little consequence for students, but our data suggest that teachers‘ relationships with one another, and their trust in the principal, cannot be easily disaggregated.

  4. While public policy and community opinion increasingly put pressure on principals to improve student performance, it is equally important to expect that principals also take actions that support instructional and shared leadership which lead to improved student learning. Increasing teachers‘ involvement in the difficult task of making good decisions and introducing improved practices must be at the heart of school leadership. There is no simple short-cut.

< < Previous | Next > >


54. Witziers, Bosker, & Kruger (2003).

55. Cohen, Raudenbush, & Ball (2003).

56. Wayne & Youngs (2003).

57. Smith, Desimone, & Ueno (2005).

58. Wahlstrom & Louis (2008).

59. Brophy (1986).

60. Wiske (1998).

61. Fenstermacher & Richardson (2005).

62. Allington (2001); Knapp (1995); and Taylor, Pearson, Clark, & Walpole (2000).

63. Kirschner, Sweller, & Clark (2006).

64. see, for example, Newmann & Associates (1996).

65. Wahlstrom & Louis (2008).

66. Wilson & Peterson (2006).

67. Newmann & Associates (1996).

68. Marzano et al. (2005).

69. Hallinger (2005); Mosenthal, Lipson, Torncello, Russ, & Mekkelsen (2004).

70. Camburn, Rowan, & Taylor (2003).

71. Stein & Nelson (2003).

72. Leithwood (2001); O'Donnell & White (2005).

73. Burch & Spillane (2003).

74. Spillane, Hallett , & Diamond (2003).

75. Halverson, Grigg, Prichett, & Thomas (2007); Silins & Mulford (2004).

76. Anderson (1998); Malen (1994).

77. Leithwood et al. (2007); Leithwood et al. (2008); Mayrowetz & Smylie (2004); Spillane, Halvorson, & Diamond (2004).

78. Marks & Louis (1997); Smylie, Conley, & Marks (2002).

79. Pounder (1999).

80. Chrispeels, Castillo, & Brown (2000); Marks & Printy (2003).

81. Harris (2009).

82. Marks & Printy (2003).

83. Serva, Fuller, & Meyer (2005).

84. See, e.g., Bryk & Schneider (2003); Hoy & Sweetland (2001); Louis (2007b); Tarter, Bliss, & Hoy (1989); Tschannen-Moran (2004).

85. Daly & Chrispeels (2008).

86. Driscoll (1978).

87. Louis (2006).

88. Wiley (2001).

89. King & Newmann (2001); Louis & Marks (1998); Smylie & Wenzel (2003).

90. Scribner, Sawyer, Watson, & Myers (2007); York-Barr & Duke (2004).

91. Hord & Sommers (forthcoming); McLaughlin & Talbert (2002)

92. See, e.g., Bryk, Camburn & Louis (1999), and Louis & Marks (1998).

93. Marks, Louis, & Printy (2002).

94. Bryk & Schneider (2002); Cascadden (1998); Friedkin & Slater (1994); Goddard, Sweetland, & Hoy (2000a); Harris (2002

95. Leech & Fulton (2008); Leithwood & Jantzi (1999a).

96. Wahlstrom & Louis (2008).

97. The method of survey administration, which involved filling out surveys during a faculty meeting, makes a completely accurate response rate difficult to determine, largely because of incomplete staff lists at the building level. In addition, a few schools that participated in 2004 dropped out for 2008, and were replaced. Because we use data from both surveys, our N of schools is thus reduced to 106 when missing achievement data are factored in.

98. Newmann & Associates (1996).

99. Louis & Marks (1998).

100. Wahlstrom & Louis (2008); Louis & Marks (1998).

101. Bentler & Bonett (1980).

102. Bentler (1990).

103. The RMSEA is .45, which is considerably higher than the suggested value of .05.

104. Hargreaves (2001); Leithwood & Beatty (2007); Little (1996); Zembylas (2003).

105. Spillane (2005); Stein & Nelson (2003).